Scoring function

This is what the doc says for the scoring function:

Submission scores for a single level are computed as follows:

level # * (10,000 - tokens used) * score multiplier

For example, if you used ChatGPT to defeat level 3, and it took you 90 tokens, your score for this level would be (10,000 * 3 - 90) * 2.

My question is, in the example, shouldn’t be 3 * (10,000 - 90) * 2 ? Otherwise the definition is wrong. Can you fix it or explain what is going on?

one more question, which specific model does “ChatGPT” refers in this context? da-vinci, gpt3.turbo or both?

You are correct, I updated the example. turbo=ChatGPT

I don;t see this scoring updated, and I’m in fact having an issue related to scores.

I finally got some api keys to run with gpt-3 models, I was able to break some levels with davinci model, but I don’t see the boos in score I expected… is there something else I should consider before submitting? scores are more or less the same than when only using flann model but, as I understood, using davinci should boost scores x2 or similar.

for instance, I’m getting 9972 for level one: AIcrowd | HackAPrompt 2023 | Submissions #228558

Please correct me if I’m wrong